Outcome
Make Zi's core load, unload, parallel-update, initialization, and plugin-authoring paths report truthful status and satisfy their published lifecycle contracts.
Context
A read-only audit of z-shell/zi next at 795662792e70af6cd36155b019b775796a9fdbee found four reproducible false-success or skipped-work defects, three substantial incomplete paths, and contract drift between Zi and the canonical Zsh Plugin Standard.
The repository's broad checks are green, but the affected paths have no focused regression coverage. The audit ran 60 existing behavioral scenarios, a 31-file Zsh syntax and compile sweep, Trunk across 67 files, and isolated reproductions for the confirmed runtime defects.
Expected deliverables
z-shell/zi: make parallel updates wait, clean up, and propagate failures.z-shell/zi: make unload and update callback capabilities work for regular plugins and snippets.z-shell/zi: execute deferred atinit correctly with nocd.z-shell/zi: make object-load status explicit and prevent scheduling after failure.z-shell/zi: complete unload ownership for hooks and repeated widget loads.z-shell/zi: establish complete, retryable home initialization.z-shell/zi: modernize or deprecate the combined scaffold and remote-creation path.z-shell/zi: consolidate confirmed duplicate helpers and resolve naming drift.z-shell/wiki: reconcile Zi PMSPEC documentation and executable validation.
Each deliverable is maintained as a native sub-issue in its owning repository.
Dependencies and risks
No external blocker is known. Recommended delivery order is parallel status correctness and lifecycle callbacks, followed by atinit and object-load status, then scaffold, unload ownership, initialization, and consistency work.
The scaffold change and unload ownership model require design review before implementation because they affect public behavior and compatibility. Fixing the .zi-load-object typo mechanically would double-count errors, so its result and aggregation contracts must be separated.
Completion criteria
- Every native sub-issue is complete or explicitly descoped with a recorded reason.
- Focused regression tests cover each confirmed runtime failure.
- Zi's broad syntax, compile, behavioral, and Trunk checks pass.
- Published Plugin Standard capability claims match tested Zi behavior.
- No required work remains only in an audit report or local memory.
Next action
Start with the parallel-update and callback sub-issues, preserving separate reviewable changes and linking each pull request to its own issue.
Outcome
Make Zi's core load, unload, parallel-update, initialization, and plugin-authoring paths report truthful status and satisfy their published lifecycle contracts.
Context
A read-only audit of
z-shell/zinextat795662792e70af6cd36155b019b775796a9fdbeefound four reproducible false-success or skipped-work defects, three substantial incomplete paths, and contract drift between Zi and the canonical Zsh Plugin Standard.The repository's broad checks are green, but the affected paths have no focused regression coverage. The audit ran 60 existing behavioral scenarios, a 31-file Zsh syntax and compile sweep, Trunk across 67 files, and isolated reproductions for the confirmed runtime defects.
Expected deliverables
z-shell/zi: make parallel updates wait, clean up, and propagate failures.z-shell/zi: make unload and update callback capabilities work for regular plugins and snippets.z-shell/zi: execute deferredatinitcorrectly withnocd.z-shell/zi: make object-load status explicit and prevent scheduling after failure.z-shell/zi: complete unload ownership for hooks and repeated widget loads.z-shell/zi: establish complete, retryable home initialization.z-shell/zi: modernize or deprecate the combined scaffold and remote-creation path.z-shell/zi: consolidate confirmed duplicate helpers and resolve naming drift.z-shell/wiki: reconcile ZiPMSPECdocumentation and executable validation.Each deliverable is maintained as a native sub-issue in its owning repository.
Dependencies and risks
No external blocker is known. Recommended delivery order is parallel status correctness and lifecycle callbacks, followed by
atinitand object-load status, then scaffold, unload ownership, initialization, and consistency work.The scaffold change and unload ownership model require design review before implementation because they affect public behavior and compatibility. Fixing the
.zi-load-objecttypo mechanically would double-count errors, so its result and aggregation contracts must be separated.Completion criteria
Next action
Start with the parallel-update and callback sub-issues, preserving separate reviewable changes and linking each pull request to its own issue.